Home > Chemical > Data Science: Collaboration Required

Data Science: Collaboration Required

Despite a great deal of hype in the technology and business press about AI changing the way organizations do business, there have been precious few companies that have actually been able to enable people from the bottom-up to leverage data science and machine learning day-to-day.

The challenge comes down to several factors, the first of which is often the most difficult to overcome: the siloed nature of business data. Accessing and aggregating data from across the organization is rarely simple, and both data scientists and analysts alike can encounter complexities with regards to compatibility, data cleaning, and de-duping, for example.

Then, of course, there’s the actual process of drawing out insights from data to turn it into business value. The most sought-after data scientists, analysts, and data team managers know that it’s not enough to play with data in a sandbox; the value comes from operationalizing projects, or pushing them to production to bring real results in terms of increased revenue, decreased costs, etc.

But it’s not enough for data professionals to simply operationalize models and projects in an ivory tower. One of the most critical pieces of the puzzle — but often the most difficult for many data scientists and analysts — is working with the business. This includes not only working alongside these subject matter experts from the beginning to create models, but also once a project is complete, ensuring that business users understand the outputs they are receiving.

All of this becomes even more difficult when considering the scale at which it needs to be done to bring continual business value. Creating one model a year isn’t going to be enough – the best AI companies are the ones thinking about how to create hundreds or thousands (or hundreds of thousands) of models a year.

Implementing Enterprise AI at scale is clearly not easy, but despite this, there have been success stories in the AI space. Earlier this week, Tech Wire Asia spoke to Florian Douetteau, the CEO of Dataiku, one of the world’s leading Enterprise AI and machine learning platforms.

READ ALSO  Design Of Anchor Bolts Embedded In Concrete Masonry

One of Dataiku’s successes has been with GE Aviation (there are other case studies here), which has leveraged Dataiku to implement its own version of a self-service data system that allows teams across the organization to use real-time data at scale to make better and faster decisions.

From Silos to Self-Service: Data Transformation at GE Aviation | John Tudor, GE Aviation

Despite the hype around AI in the media, very few businesses have managed to execute on incorporating the fundamental processes that enable these data insights at scale, much less automating them to enable AI services. In this talk, Jon Tudor, Sr.

One of the reasons that GE Aviation has had such success, Douetteau articulated, is because they believe that one way to address all of the aforementioned challenges is to democratize the use of data — that is, move it from a specialized, exclusive practice to something accessible for all, no matter what their job title.

“Articulating business problems in a data domain is a real challenge for many teams and in many companies,” said Douetteau. “The solution to ensuring that data solutions are actually solving real business problems is that, in most cases, people experienced in the business domain need to step in and get their hands into data.”

One of the fundamental features of Dataiku that has been present in the product since the very beginning (the company was founded in 2013) is that the platform provides a collaborative framework where data scientists, analysts, and people from the business side can work on data projects together. This is the foundation of democratizing AI efforts in an organization.

Of course, technology professionals at management level right up to the C-suite have been investing in low-code or no-code tools since the beginnings of business intelligence (BI). But when it comes to AI, scaling is about more than just people and technology, it’s also about adding efficiency into processes.

READ ALSO  Sahara Group Tasks Stakeholders On Africa’s Energy Transition

“There is no such thing as a magic bullet solution when it comes to AI, not with people (i.e., hiring a bunch of data scientists), and not with technology” said Douetteau. “But one thing that is really, really important is automation. That’s one part that Dataiku makes easier: automating 80 percent of the work in a data project, repeatedly.”

That last word is an important one in the context of data processing: there’s no easy button that unearths meaningful insights in a business context. There’s still a large amount of hard work and trial-and-error, even with a platform like Dataiku doing a lot of the heavy lifting.

Dataiku: Your Path to Enterprise AI

null

The emergence of the capabilities of the Dataikuplatform is good timing. At present across APAC (and indeed across the world), there’s a significant shortage of qualified data scientists. Of those who are on the job market, most are young graduates (it’s a young discipline) with little practical business experience. Dataiku essentially provides a way in which those steeped in a business as analysts, can — with a little help from one or two data scientists — develop powerful data projects in their own right.

On the other end of the spectrum, to be successful in AI endeavors globally, companies must provide data scientists with the tools they need as well that make them more efficient across the entire data lifecycle (data exploration, coding, connectivity, automation, model deployment, machine learning, etc.). Reuse and automation is really the name of the game and is key to scaling, which means it must be a cornerstone in tools for both technical and non-technical people at the company alike.

  1. Source: techwireasia

Total Views: 124 ,
0
0

Leave a Reply

Your email address will not be published. Required fields are marked *